reaatech/structured-output-repair

★ 0Last commit: May 12, 2026GitHub →

These packages provide a pipeline to automatically fix malformed LLM outputs, such as invalid JSON, markdown fences, or type mismatches, to ensure they conform to your defined schemas. You would adopt them to prevent agent system crashes caused by unreliable model responses. The system uses a graduated, multi-strategy repair engine that can be integrated directly into your TypeScript code via `@reaatech/structured-repair-core` or exposed as an MCP server for use in tools like Claude Desktop.

agentic-ai ai ai-agents anthropic developer-tools json llm openai typescript

Packages

Sort

2 packages

structured-repair-core

@reaatech/structured-repair-core

pending npm

Fixes malformed LLM JSON output by applying a configurable pipeline of repair strategies to ensure compatibility with a provided Zod schema. It exports utility functions that return validated, type-inferred data or detailed metadata about the repair process.

View package

status: awaiting publish

structured-repair-mcp

@reaatech/structured-repair-mcp

pending npm

Exposes tools for repairing malformed LLM outputs against JSON schemas via the Model Context Protocol. It provides an MCP server that can be run as a standalone process or integrated programmatically to analyze and fix JSON syntax, type mismatches, and formatting errors.